Please try to imagine. The data-intensive world of the future will be powered by campuses of hyperscale data centers dedicated to managing capacity demands. Is it a bizarre dystopian prediction or the future of hyperscale data centers?
Hyperscale facilities are more than 5,000 servers with a minimum size of 10,000 square feet, provide at least 40MW of IT capacity, and are typically identified as data centers serving enterprise customers.
scale
From the smallest edge sites to large colocations and multi-building campuses, hyperscale facilities are the largest standalone sites available. So, logically, the next iteration of the data center will continue to scale linearly to meet increasing data demands.
Darren Watkins, managing director of Virtus Data Centers, talks about the scale of hyperscale's next enterprise data center facility: But now, with the overlay of artificial intelligence (AI)/machine learning (ML) adoption, demand is increasing even faster in all regions.
“While locations have been very specific and carefully chosen over the last few years, new AI/ML workloads are often latency-insensitive and power-constrained in large metropolitan centers. This opens up new locations as potential data center locations. The cost of installing a facility is disproportionate to the IT workloads that location can support, requiring sites of scale and accessibility. ”
Taking a slightly different perspective, Terry Storrar, managing director of Leaseweb UK, believes that land scarcity and power requirements will impede this linear growth in data center scale.
“Like any building project, the size of a data center depends on adequate land and utilities to make construction viable,” he says.
“A global shortage of available power will limit size. Despite rumors of some mega-projects in Europe this year, it is possible that data center scale will continue on a linear trajectory in the long term.” Sexuality is low.”
Jon Healy, COO of Keysource, says there are two sides to this question. The first is technology. Because data centers are driven by the technology deployed inside them.
“We're going to start to see a shift in the types of computing and storage technologies that users and service providers are adopting, and that's ultimately going to kind of impact the data center landscape,” he says.
“We're now evaluating the types of technologies people are using based on the types of workloads they're running and starting to segment those workloads and figure out which technologies are best for them. And we're starting to segment out what environments are probably best suited for, which data centers may be best suited to support that type of technology itself.”
The second part, Healy said, is the number of global resources available to accommodate these hyperscale data centers.
Consistent with Strahl's points about land and power shortages, he believes more red tape will likely be added to protect these assets.
“I think we're going to see more constraints and more resources being protected geographically for the data center industry to use these resources,” he says.
“I think we will see more zoning regulations introduced, such as through planning application approvals and higher standards, which if data center providers cannot meet will result in them not being given the land and power to provide their services. ” It’s their business. ”
difference
So what do we expect the next hyperscale facility to look like and how will it be different from what we have?
Anthony Milovantsev, partner at Altman Solon, predicts that these facilities will be shaped by eight major trends in future innovation.
First, he says, “we have certain dedicated rooms and products with direct chip cooling or full immersion liquid cooling.”
Second, it improves modularity and reduces time to market. Third, there will be an inter-campus connectivity fabric.
On the environmental side, sustainability elements such as rooftop solar panels will be added.
Additionally, you'll be giving back to the local ecosystem, such as by pumping waste heat to local cities and towns. There is also the potential for highly differentiated pricing and tiering to accommodate customers who do not require traditional protections or delays.
He said the company will develop “a ton” of innovation for the edge, including data centers under towers, under skyscrapers and in parking lots.
Finally, it will have new futuristic features such as underwater data centers, quantum computers, and even a data center on the moon. However, he hastens to add, “That's still more than 30 years away.”
sustainability
Healy talks about these new requirements feeding into the sustainability part of future hyperscale data centers. That's because he believes “we're going to start to be evaluated on how we're delivering these hyperscale or any type of data centers in a more sustainable way.” To get those resources. ”
When you think about the power and cooling demands of this upcoming data center, you wonder how it will balance with sustainability requirements.
Shareef Alshinnawi, vice president of strategic accounts at Iceotope, says optimal efficiency is the key to solving this challenge.
“Between increased energy usage, rising power costs, and potential government regulations, data center operators are under pressure to reduce the energy consumption of their facility cooling systems.”
“Addressing the growth in computing while reducing energy and natural resource consumption requires alternative power and cooling technologies, such as precision liquid cooling. Given the scale of data centers and their infrastructure, systems and cooling Efficiency in design provides dramatic benefits in reducing consumption. It’s out of scope.”
Efficient power and cooling is not always easy in any type of facility. Mark Grindey, CEO of Zeus Cloud, said: “Older data centers may have limitations in terms of power and cooling, as older equipment may be expensive to upgrade or may not have the space or capacity to do so. If it’s not built with new technology in mind.”
However, all is not lost, with some operators choosing to build solar farms and wind turbines to offset their carbon emissions and electricity usage, but not necessarily at the data center itself. data centers are still sustainable.
Another thing data centers do to become more sustainable is to recycle the air and water used to cool data center halls. ”
A.I.
AI is still in its infancy, only a small portion of the population is using it, and the technology is still evolving, so who knows how big the capacity needs will be in the long term? I don't know either. Given the acceleration of the AI market, will hyperscale facilities be sufficient to meet future data demands in the near term? If so, for the long term?
Pete Overell, Panchaea's founder and sales director, said, “Hyperscale facilities can support AI for now, but computing demands are exploding.”
“Data centers will need access to hardware and hosting solutions to meet these demands in the medium to long term. Forging strategic partnerships now will help meet this challenge in the future.
Matt Rees, CTO of Neos Networks, echoed this sentiment:
“Thus, we may see people replacing their existing compute with more powerful devices. For example, we continue to optimize OpEx spending, much of which is rack space and power in commercial data centers. So how can you use a smaller form factor and still be able to offer the same high-capacity, high-availability connectivity products to your customers?”
Spencer Lamb, CCO at Kao Data, puts his position on this a little more directly: Overcoming production capacity shortages within the market. ”
He says this is an issue that needs to be addressed at both an industry and government level. In the UK in particular, we need to consider how the country's critical infrastructure is mapped and provisioned.
“Additionally, data centers will need to be designed and, in some cases, modernized and redeveloped to accommodate high-density computing, increasing countries’ technological resilience and readiness for the next wave of AI. “We need to establish a secondary hub in the region,” he said. I would add.
corner
Given this, one wonders whether we will reach a point in the future when we will need smaller rather than hyperscale facilities, even outside of edge environments, given the complexity and expense associated with building a data center. Some people may think so.
Overall, he disagrees and says that “smaller data centers will be essential going forward. These will be a big factor for hyperscalers to remain agile and flexible. Regional data processing will be , especially until hyperscale data centers become more viable in dense urban environments. Being able to scale up and down in smaller facilities is extremely beneficial.”
Additionally, Grindy said small data centers are also necessary for startups, small businesses and enterprises, as they allow companies to deploy their own kit and ensure that security protocols and data policies are met. It reminds me of.
“While small data centers are important, they may start to decline in the future as they can be difficult to keep up with in terms of technology. “Only mid-sized data centers that can do that are likely to emerge,” he says.
None of us have a crystal ball, but thanks to technologies like AI and quantum, this next phase of data center evolution promises to be interesting.
But should we call this a so-called hyperscale 2.0 facility? Is it megascale, terrascale, petascale, exascale, zettascale, or Yottascale as proposed by Kao's Ram? Personally, I would like to invest in Gigascale. is being introduced.